全球智库信息集成服务系统

Model Results Will Differ Based on User Inputs

• The model uses the following user-specified inputs: the growth in vehicle miles traveled in a baseline future without HAVs, the change in safety rate of non-HAVs, the timing and extent of HAV diffusion, the impact of HAVs on transportation demand, the safety performance of HAVs initially and over time, and the upgradeability of HAVs.
• Illustrations of model use show that the future may unfold in complex and important ways that are not obvious when attention is paid only to safety at the time of HAV introduction. Such information might raise different policy questions, such as how policies can be crafted to encourage rapid learning or how trade-offs should be made between smaller short-term risks that may lead to larger longer-term gains.

Highly automated vehicle (HAV) safety is a principal concern for the transportation industry, policymakers, and the public. Much of the concern is focused on the question of how safe HAVs should be before they are allowed on the road for consumer use. However, the safety performance and impact of HAVs will change over time and may have much greater effects in the long term than in the short term. Thus, to address this concern properly, stakeholders must also ask how safe HAVs will become over time and how policy choices made today could shape the future of road safety.

To help answer these questions, this report describes a simple model of how factors shaping HAV diffusion and performance may interact and result in different safety outcomes over time. It can be used to measure safety in terms of injuries, crashes, or other metrics. Importantly, the model does not predict the pace of future HAV diffusion, the rate of change in safety performance, or other factors. Instead, given a user's hypotheses, it estimates how many lives would be lost each year in a future with and without HAVs. The model helps users explore how different assumptions lead to different outcomes and which factors may lead to greater or lesser safety. Such insights can help users consider how policies might shape HAV diffusion and performance to improve safety now and over time.

• Chapter One

  Introduction

• Chapter Two

  Problem-Framing and Definitions

• Chapter Three

  Modeling Safety in a Future Without Highly Automated Vehicles

• Chapter Four

  Modeling Safety in a Future with Highly Automated Vehicles

• Chapter Five

  Model Exploration and Illustration

• Chapter Six

  Model Summary and Conclusions

• Appendix A

  Modeling Diffusion with a Logistic Function

• Appendix B

  Modeling Learning with an Exponential Decay Function